TWI220505B - Image enhancement method - Google Patents

Abstract

The present invention is related to an image enhancement method for removing spots of skin color in the image. At first, the format value of the second image for each pixel in the image is calculated. Then, whether the pixel is located on the boundary is judged for each pixel so as to judge if the second image format value of the image is within one skin color range. If the pixel is not located on the boundary and the second image format value is not within the skin color range, a spacial smooth filter mask is dynamically generated, and a smooth process is then conducted onto the pixel based on the spacial smooth filter mask. Finally, a spacial smooth filter mask is dynamically generated for each new pixel in the new image after it is undergone with the smooth process, and the entirely smooth process is conducted onto new pixel based on the smooth filter mask.

Description

1220505 V. Description of the invention (l) The present invention relates to an image enhancement method: and particularly to a method that can beautify the skin tone in an image. With the trend of digitalization of images and the popularization of image capture equipment, such as digital cameras and scanners, many image processing methods or special effects have been developed to edit the images used. &lt; / In the method of image enhancement, for example, people in addition to the central eight are usually using a smooth wave device / directly to prevent the red, green, and toilet channels in the image from within the range of color difference Pull. This smoothing process makes the effect of the conventional image improvement method =: noise removal. When there are only a small number of noise points, it can be &gt; depending on the number of noise points, and when the noise points ^ are distributed on ^, the expectation of eliminating black noise can be achieved; the noise is averaged, but cannot It is true that the effect achieved by the second method is another purpose of the image improvement method, which is to eliminate noise. For each image in the entire image 3 === The disadvantage is that, because the smoothing process will be blurred, faded, or even processed, the boundary area in the image is also in view of this. The noise on the main color of the present invention is important for smoothing. The processing method provides a beautified image that can remove the perfect appearance of the skin part = β, and the method of removing the skin color in the image from 糸 7, 〇 ·, 丄 VP m〒 skin color is to achieve the above purpose of the invention. Strengthening methods to achieve. The error is achieved by the two-stage smoothing method proposed by the present invention. In January, the image enhancement method is profitable. The first stage of the smoothing process is 0599-6516W; 200M7; Yianhou.ptd Page 4_ 1220505 V. Explanation of the invention (2) The main purpose is to drastically reduce 炝 a. The main purpose is to eliminate the spots in the needle skin. The main purpose of the second stage of smoothing is to make a further modification of the entire daytime surface. The image is :: ϊ ί :, First, • Receive a value in the first image format 彳 111: The second image format value for each pixel in the image. However, the first of the pixels is to determine whether the pixel is located on the boundary, and whether the image format value is within the skin color range value. Within range 1, m: on the boundary and the value of the second image format is not in the skin color gate J} L 冰清 # 你 s ~, a spatial smoothing filter mask is generated, and according to this space, a thousand β j wave mask pair Pixels are smoothed. Mai z :: t zhongzhong 'To ensure that this space smoothing filter can be used to mask A &amp; π m noise, the establishment of this space smoothing filter mask is to increase the weight of this pixel &amp; The surrounding, located within the range of skin color values, inflame this non-honing = to use these neighboring pixels located within the range of skin color values = pixels within the range of p iff. In other words, # 此 pixel 四 _ i Axi, the neighboring pixels in the skin color range (indicating: this pixel's neighbors go in 'bei 1 this pixel is strengthened with reference to these bits after smoothing within the skin color range value , Its color will be faded out. There are almost no neighboring pixels in the skin color range around λ other than the V color in the skin color), then this pixel will not refer to these pixels that are not in the 2 range. The color will not be changed. Among the new images after the last smoothing &amp;#I 料 # #this, a smoothing filter mask is generated, and the new pixels are overall smoothed according to the smoothing filter mask. Brief description of the drawings Page 5 0599-6516TW; 200M7; Yianhou &gt; ptd 1220505 V. Description of the invention (3) In order to make the above-mentioned objects, features and advantages of the present invention more comprehensible, the following examples and examples are given in conjunction with The drawings are described in detail as follows. FIG. 1 is a flowchart showing a method of eliminating noise in an image enhancement method according to an embodiment of the present invention. Figure 2 is a schematic diagram showing eighth values of skin tone, saturation, and brightness (Hsv). 77 Figure 3 shows an example of a type of spatial smoothing filter mask. Figure 4 shows an example of another type of spatial smoothing mask. FIG. 5 is a flowchart showing an overall smoothing processing part of an image enhancement method according to an embodiment of the present invention. Explanation of symbols S100, ..., S600, S700, ..., S900 ~ operation steps; 1 00, 4000 ~ spatial smooth filter mask; 2 0 0 ~ HSV value mask; 3000 ~ distance mask; 1100, 2100 ~ pixels. Embodiment Next, referring to the figure, the method for beautifying the skin color in an image according to the embodiment of the present invention will be divided into a noise removing portion (the first stage) and an overall smoothing portion (the second stage). <Removal of Noise Points> Referring to FIG. 1, FIG. 1 is a flowchart showing the removal of noise points in the skin color in the image enhancement method according to the embodiment of the present invention. First, in step S100, a red, green and blue (RGB) (first image) is received.

0599-6516TWF; 2001-17; Yianhou.ptd Page 6 1220505

Format). Then, as in step S200, calculate | once, you can adjust the tone, saturation, and brightness (color of one pixel in HSVK).

Alas. It is profitable to calculate the HSV value in the second step (convert the HSV value to the second grid dimension value. ㈣ Change the a formula or conversion matrix to directly convert RGB (edgfl, 'for each pixel, use the detect edge (wide detect) method) Judging whether the pixel is 340 〇 / Γ A schematic diagram of the distribution of skin color values, where 0 ~ 20 and = ~ 〇 are the range of the threshold value of general skin color (that is, _2〇 ~ + 2〇), 3 values of HU-like skin color The range interval, 0.02 ~ 0.75 is the general skin value i &amp; m. In other words, the step S4QQ is to determine whether the η Τ of this pixel is between 0 ~ 20 or 340 ~ 〇, and whether the S value is 〇. ^ 0.5 疋Whether it is between 0 · 02 and 0.75. 非 m Non-if If r = The pixel is not located on the boundary and the pixel's HSV value is not within the skin color envelope value, then a filtering mask is dynamically generated as in step S500. . This spatial smoothing filter mask is used to remove the skin color; 丄 'moon point, the method of establishing it will be described later. "Then, according to step S600, smooth the pixel based on the dynamically created spatial flat mask 4 # All pixels in the image = children * ^ after the judgment and smoothing of steps S300 to S600. The work points &amp;. 4 will ① to purify the solid point of the step S500 and S60 0 may be at least two different patterns of

0599-6516TW; 2001-17; Yianhou.ptd Page 7! 22〇5 05 5. Description of the invention (5) In the first type, a method for dynamically generating a spatial smoothing filter mask (/ step S500). First, an empty spatial smoothing mask is established, and then, it is judged whether the HSV value of neighboring pixels around this pixel is within the skin color range value. If the HSV value of the neighboring pixels is within the skin color range value, the position record corresponding to the neighboring pixels in the spatial smoothing filter mask is valid and given a fixed weight (such as: 1/9, 3X3 spatial smoothing filter mask The cover is ... for example). Fig. 3 shows an example of a spatial smoothing filter mask created according to the method of establishing a spatial smoothing filter mask of the first type. This spatial smoothing filter mask 1 00 corresponds to a mask of a pixel 1 i 00, and the values in parentheses in the pixels represent their corresponding Η values (to simplify the diagram, the s and v values are omitted) . In this example, if the value of the neighboring pixels around the pixel 丨 丨 〇〇 is between 0 ~ 20 or 340 ~ 0 (or -20 ~ + 20), then the space smoothing filter mask 1 00 0 corresponds to this The position records of neighboring pixels are valid for, and ν), and given a fixed weight (such as: 1/9, 3X3 spatial smoothing filter masks are used as, and, for example); and if the neighboring pixels around pixel 1100 are If the value of Q is not between 0 ~ 20 or 340 ~ 0, the position record corresponding to this in the space smoothing filter mask 1000 is,, N (invalid), and a weight of 0 is given. Profit = method to set a spatial smoothing filter mask, to ensure that only neighboring pixels located in the skin color range will compensate for pixel 1 100. Next, a description will be given of a method of performing smoothing processing S60> 0) corresponding to the first type. First of all, the RGB values of neighboring pixels corresponding to all the positions in the spatial smoothing filter mask that are valid (the weights are as recorded in the gate, sub,, and filter masks) and 1 ^ (1 ^ ( ^ Value (1 = slip 丨 its resistance hook i minus

1220505 V. Description of the invention (6) A weighted operation is performed on the weights of adjacent pixels that are recorded as valid, to obtain a weighted pixel content value. Then, replace the RGB value of this pixel with the weighted pixel content value. Taking Figure 3 as an example, the new RGB value of the pixel is the spatial smoothing filter mask. The RGB values of the neighboring pixels corresponding to the positions of six records in the 1000, V, and V &quot; are weighted as: 1/9, and the RGB value of the pixel 1100, its weight is 3/9 = Bu 6 * (1/9), the weighted result. With this method, around pixel 1 100, only neighboring pixels within the skin color range value will compensate pixel 1 100. When the number of pixels 丨 i 〇 〇 around, within the range of skin color, the more pixels (indicating that the pixel 1100 is the higher the probability of noise in the skin color), the smoother it will be closer to the skin color. Conversely, when there are very few pixels around the pixel 1100 within the skin color range (indicating that the pixel 1 100 is a low probability of noise in the skin color), the smoothing process will be very close to the original color. In addition, in the second type, a spatial smoothing filter mask is dynamically generated (step S500). First, create an empty spatial smoothing mask, and then set this based on the distance between the neighboring pixels around this pixel and the target skin color (or the correlation between HSV values) and this pixel and its neighboring pixels. The weight of the position of the spatial smoothing filtering mask corresponding to neighboring pixels. Fig. 4 shows an example of a spatial smoothing filter mask created according to the method of establishing a spatial smoothing filter mask of the second form 4 00. This spatial smoothing wave mask 4 00 is a mask corresponding to a pixel 2 1 0 0, and the value in the brackets in the pixel represents its threshold value (to simplify the diagram, the s and V values are omitted). In this example, according to pixel 21 〇 and its surrounding neighboring pixels and _

1220505 V. Description of the invention ⑺ &quot; 1 The difference between the HSV value between the target skin color (such as Η value equal to 0) and the distance between the pixel 2100 and its neighboring pixels can generate an Hsv value mask 2000 and a distance mask respectively. 300. In this example, the neighboring pixels that are closer to the target skin color have higher corresponding weights in the Hsv value mask 2000 (eg, the weight value between the value of η and the target skin color is within 10, and the threshold value is equal to the target skin color. The weight with a difference within 10 ~ 20 is 1 ', and the weight with a difference of more than 20 from the target skin color is 0); the weight of the pixel 2100 is a fixed value, such as 3. In addition, the closer the pixel 2 100 is, the higher the weight value in the distance mask 3000 (eg, the weight value of the pixel 210Q is 2 and the diagonally adjacent pixel 2100 is 1); The weight of 2100 is a fixed value, such as 3. The combination of the HSV value mask 2000 and the distance mask 3000 (for example, multiplication) results in a spatial smoothing filter mask 4000. In other words, 'the closer the pixel 2100 and the closer the pixel to the target skin color, the higher the degree of compensation for the smoothed pixel 2 100. The weights recorded in the spatial smoothing mask 40 0 0 can be used for subsequent smoothing. Next, a method for performing a smoothing process (step S60 0) corresponding to the second type will be described. First, filter all records in the mask based on spatial smoothing

The weight value 'performs a weighted average operation between the corresponding RGB values of these neighboring pixels and the RGB value of the pixel 2100 to obtain a weighted average pixel content value. This pixel's RGB value is used to weight the average pixel content. Value replacement. <Overall smoothing processing section> Next, referring to FIG. 5, FIG. 5 is a flowchart showing an overall smoothing processing section in an image enhancement method according to an embodiment of the present invention.

1220505

^ ^ O i UU A new pixel-per-pixel smoothing process: σ1 :: 'π s_ in the image, for each new image 2 in the new image: step mask, and finally, step S900, According to this smoothing and smoothing filtering, the overall smoothing is performed ... when the new shadow mask is applied to this new pixel. With the overall smoothing of S900, H = prime the work of step part. The entire smoothing process is performed in the same manner as described above in the steps S800 and S90. 0 can also be at least two different. In the first type, a smooth filter-empty smooth filter mask is generated. Then, "new": the method is to first establish the degree of difference between the HSV value and the new pixel. When the position weight of the neighboring pixels in the neighboring pixel mask and the smoothing filter are not determined, the new RGB values and recorded weights of these neighboring pixels at the overall smoothing according to this smoothing filter are used. The corresponding operation is performed, and the surname and RGB values of the weighted average operation are weighted and averaged. In the second type, a new value of flat pixels is substituted for the new pixel. The new method uses the conventional smoothing method. The new pixels perform overall smoothing. Therefore, the method proposed in the present invention can use two-stage smoothing to remove the effect of beautifying the skin color in the image. Skin in the image = skin spots on the skin for smoothing, although the present invention has been presented in a better part. Limiting the invention 'anyone familiar with this ^ ^ as disclosed above', but it is not intended to And supporters, without departing from the essence of the present invention

1220505

0599-6516TW; 2001-17; Yianhou.ptd Page 12

Claims (1)

1220505 VI. Application for Patent Fanyuan1 · An image enhancement method, receiving the first-image grid method includes the following steps: For the n in the image and the age I: pixels, perform the following steps: Determine whether the pixel is located on the boundary; determine whether the pixel's _shang-dan / μ μ_ is not within the range of an image format; if the pixel is not located on the boundary and the first: non Within the skin color range value,!, Dynamically generates a spatial smoothing U = J and performs a smoothing process on the pixel according to the spatial smoothing mask: 2 · As described in the first item of Shen Qing's patent, It also includes calculating the first image format value of the pixel 4; 3. The method described in the first item of the patent application, which further includes the following steps: For each new pixel in the image after the smoothing process, a new image For each new pixel, the following steps are performed: generating a smoothing filter mask; and performing an overall smoothing process on the new pixel according to the smoothing filter mask. 4 · The method described in the first item of the patent application, which The first image format value is a red, green and blue (R (JB) value. 5) The method as described in the first item of the patent application, wherein the image format value of the second scene is the hue, saturation, and brightness (HSV) values. 6. The method as described in the first item of the patent application, wherein determining whether the pixel is located on a boundary is performed by using an edge detection method. ! 22〇5〇6. Patent application scope 7. If the application of the patent smooth filtering mask is established, an empty space is judged to be within the color range of the pixel; and if the adjacent pixel is within, the space is recorded as valid . 8. If applying for a patent, the smoothing method sets a position where a fixed weight is valid; using the spatial weighting value, calculates the pixel, performs a weighted operation on the value of the neighboring pixel that smoothes the space, and generates the generation of the pixel. 9 · The method described in the patent application for creating a smooth filtering mask according to the second image grid item between the pixels, wherein the method of dynamically generating the empty method includes the following steps: smooth filtering mask; proximity Whether the second image format value of the pixel is within the skin's first image format value in the skin color range sliding filtering mask corresponding to the position of the adjacent pixel, the method described in item 7, wherein the pixel is performed, It includes the following steps: All the records of the spatial smoothing filter towel are weighted to the positions where all records in the sliding filter mask are valid; all the records of the filter mask towel are valid positions and the image format value and the pixel The first image format to obtain a weighted pixel content value, and the first image format value to take the weighted pixel content value to take the method described in item 1 wherein the dynamic generation of the empty method includes the following steps: Mask, and the difference between the second image format value and the -target value expression value, the degree of the pixel's neighboring pixels and the ! 22〇5〇6. The scope of the patent application. The prime distance determines the weight of the position of the space-time taxiing wave cover. &Lt; The method for performing the smoothing process includes the following steps: According to the weights of all records in Wen Erjian's smooth filter mask, Ding Shang will take the value of the-image format of the neighboring images * and the image: J value- The weighted average operation is performed to obtain a weighted average image; = value, and the value of the first image format of the pixel is replaced by the internal value of the weighted average pixel. Let the target value be one of the smoothed filter masks. Among them, the smoothing filter 11 is generated. • The predetermined skin color value is determined by the method described in item 9 of the patent application. 1 2 • All positions in the method cover as described in item 3 of the patent application are recorded as valid. 1 3 • As applied The method of the method wave mask described in the third item of the patent includes the following steps: establishing an empty smooth filter mask; and setting the smoothing according to the second image mode value difference and distance of the new pixel and its neighboring pixels. Correspondence in filter mask The position of the neighboring pixel is ^ two weights. 14. The method according to item 12 or 13 of the applied patent, wherein the method for performing the overall smoothing processing on the new pixel includes the following steps. The weights of all records in the smooth filtering mask will wait for the new image-value of the neighboring pixel and the new image of the new pixel. Page 15 0599-6516TWF; 2001 -17; Yi anhou.ptd 1220505 6. Perform a weighted average operation on the format value of the patent application to obtain a weighted average pixel content value; and replace the new first image format value of the new pixel with the weighted average pixel content value. * ❿ 0599-6516TWF; 200M7; Yianhou.ptd Page 16